Analysing Multiple Explosions in Europe:

Page 1British Crown Copyright 2008/MOD

Analysing Multiple Explosions in Europe:

Signal Characteristics and Propagation Modelling

D. Green, J. Vergoz, A. Le Pichon, L. Ceranna, D. Drob and L. Evers


Analysing Multiple Explosions in Europe:

Signal Characteristics and Propagation Modelling

D. Green, J. Vergoz, A. Le Pichon, L. Ceranna, D. Drob and L. Evers

Summary• The 4 Explosion Series

• Identifying Multiple Explosions

• Propagation Modelling

• Explosive Yield Estimation• Discussion / Conclusion

• Observations

• Gerdec Explosion


The Explosions

Explosion

IMS array(operational)

Proposed IMSarray

Non-IMS array


Detections: F-Detector

Beam

Filtered( 0.3-0.8 Hz)

F-Value

Probability

(SNR = 2)

• Detections made using the F-detector (e.g., Blandford 1974)

• Optimal detector for perfectly correlated signal in Gaussian, stationary noise

• In presence of signal, F has a noncentral probability distribution – can assign a probability that an F value is a detection at a given SNR.

I26 Detections from Chelopechene


Detections: F-Detector

Beam

Filtered

(0.3-0.7 Hz)

F-Value

Probability

(SNR = 5 )

BNIA Detections from Chelopechene

• Useful for picking signals out of a noisy background

• Example below: signals at a range of 2097km recorded at BNIA (Blacknest)

• Small aperture array (~200m), lots of cultural noise (roads)


Explosion Timings

Expected Detections

To the east

To the west

• Explosions occurred at times of strong stratospheric zonal wind

• The detectibility along a given azimuth – controlled by along-path strato. wind


Observation Locations

Buncefield Chelopechene

Novaky Gerdec

• Most observations as expected – downwind of source

• Some exceptions – e.g., upwind arrival at I48TN from Gerdec

• Downwind – signals have been detected at distances of over 4500km


Spectra Comparison: IS26

IS26: Detected all four explosions

SignalNoise

• Gerdec – Power at >30s

• Novaky – Upwind, very low SNR

Note:

Yield Estimate (from period):

(AFTAC eqn e.g., Edwards et al., 2006)

(note comparison withpressure yields – p.24)


Multiple Explosion Observations

GerdecTIR ~ 17km

ChelopecheneVTS ~ 25km

Local Seismic Records

• Cross-correlate 1st six seconds of P-wave.

• Find similar seismic signals

• Assume: similar signal = similar source at similar location(λ/4 hypothesis)

= master explosion

= cross-correlation > 0.6

(all associated with above noise SNR)

Chelopechene – 7 Explosions

Gerdec – 2 Explosions


Example of Observations: Chelopechene

• Chelopechene – 03/07/2008

• All detections in westerly direction

- as expected, Nth Hemisphere summer

• Multiple arrivals from each explosion

• Overlap of arrivals from first two large explosions

• A complex set of signals to explain. - how to extract source information? - what can we tell about propagation?



• We want to: identify arrivals and compare different explosion signals.

• Simplify by smoothing out fluctuations using envelope function

• Frequency-time plot can highlight small amplitude signals (e.g., Dziewonski et al., 1969‘Multiple Filter Technique’)

Signals at I48TN from Chelopechene, Bulgaria

Narrowband Filtered (2.2-2.4 Hz)



Waveforms Log10 Smoothed Envelope

(Narrowband filter 2.2 – 2.4 to Hz applied)

Overlapped Signals – useseismic arrivals as independent timebase



I48TN I26DE

(2.0-4.0 Hz) (0.4-3.0Hz)

Exp. 2Exp. 6Exp. 7

• Exp 2 36 minutes Exp 6 49 minutes Exp 7.

• At both stations early (fast) arrivals are missing from last explosion (explosion 7).




Gerdec Explosion - Observations

• 6 infrasound observations

Backazimuths, uncorrected for wind deviations

• Observations through a wide range of azimuths



I26DE (Germany)

I48TN (Tunisia)

• Perpendicular to dominant stratospheric wind dirn.

• Upwind arrivals observed from both explosions

• Arrivals contain frequencies >1Hz and have arrival times consistent with stratospheric arrivals.

G2S

ECMWF

• Ground-to-Stratosphere waveguides are weak, or non-existent.



• Explosions separated by ~ 19 minutes

• Very little difference between explosion envelopes: - arrival times almost identical - small differences in relative amplitude


Propagation Modelling

Available tools

Major Features to explain

• Atmospheric Parameterisations

- HWM, ECMWF, ARPEGE, G2S

• Ray Tracing (1D to 3D)

• Parabolic Equation Methods (Inframap)

• Multiple infrasonic arrivals

• Potentially upwind stratospheric arrivals

• Temporal changes in waveforms

• Chebyshev Pseudospectral Method (L. Ceranna)


G2S – ECMWF – ARPEGE - HWM

ECMWFG2S

HWM

Atmospheric Parameterisations

ECMWF and G2S - very similar up to altitudes of ~30km- significant differences in the stratosphere- what are the uncertainties in T,U and V?

Gerdec to IS26 (average profiles)

ARPEGE


Gerdec – Propagation ModellingGerdec to IS26 (Germany) – PE model (J.Vergoz)

• Very weak waveguide – difficult to explain the multiple (6) arrivals from each explosion.

300 400 500Eff. Sound Speed (m/s)

120

80

40

0

Alt. (km)

120

80

40

0

Range (km)0 400 800

-20

-60

-100

dB re 1km(@2Hz)

I26DEsource

Data


Gerdec – Propagation ModellingGerdec to IS48 (Tunisia) – PE model (J. Vergoz)

• Elevated waveguide only

• Little energy scattered/diffracted to ground

300 400Eff. Sound Speed (m/s)

120

80

40

0

Alt. (km)

Range (km)500 1000

-60

-140

-220

dB re 1km(@2Hz)

source I48TN

Data

120

80

40

00


Gerdec – Propagation ModellingGerdec to IS48 (Tunisia) – Chebyshev (L. Ceranna)

• Elevated waveguides – again diffracted arrivals return energy to the ground

Data

Eff. Sound Speed (m/s)300 400

0

60

120

Alt. (km)

1000 20000

60

120

Alt. (km)

Range (km)

0

-20

-40

dB

(@ 0.2Hz)

Eff. Sound Speed (m/s)0

source I48TN


Gerdec – Propagation Modelling

• Elevated waveguides – again diffracted arrivals return energy to the ground

Data

1000 20000

60

120

Alt. (km)

Range (km)

0

-20

-40

dB

(@ 0.2Hz)

01000 2000Range (km)

0

0.24

0.28

0.32

Cel. (km/s)

• Correct celerity for observed arrivals

source I48TN

Gerdec to IS48 (Tunisia) – Chebyshev (L. Ceranna)


Propagation Issues

Gerdec to IS26 (Germany)

• How are the multiple arrivals generated?- modelled waveguide is too weak

• Would a narrow elevated waveguide generate the large number of arrivals?

• How does energy leak down to the ground surface?

Gerdec to IS48TN (Tunisia)

• Propagation upwind – energy is trapped in an elevated waveguide

• How does energy leak down to the ground surface?


Propagation Issues

Eff

. S

ound

Spe

ed (

m/s

)

Gerdec to IS26 (Germany)

2D and 3D structure• Preliminary investigations suggest that the 2D structure does not contain significant changes in waveguide structure.

Temporal Changes• Observed in data – unlikely to be resolved in models?


Yield Estimation

Whitaker scaling relation used:

• For the largest explosion of each series.

• Pressure corrected using mean along-path wind values at 50km altitude.

• Ranges at particular stations show the amplitude differences between multiple arrivals for the one explosion.

(note comparison with pressure yields –p.9)


Yield Estimation – Multiple Exp.

Chelopechene Gerdec

• Consistent differences observed between stations: - influence of : instrument calibration

wind values taken from the model?

• Only largest arrival shown on these plots


Conclusions / Further Studies

• Four large explosion series (10-1000 tons TNT) have been well recorded over the infrasound network in Europe and Asia.

• Detections predominantly influenced by the stratospheric wind direction - however propagation modelling does not always predict the correct number of observed arrivals.

• The Chelopechene explosion series suggests that explosions of down to a few tonnes of high explosive equivalent can be recorded at stations ~1000km distant from the source.

Future Improvements

• Propagation path identification- If the atmospheric models cannot explain the arrivals, what information can we use to identify the most probable path?

• Temporal signal changes- What do these tell us about the atmospheric variability?

• Downwind of the large explosions, detections can be made at over 4500km from the source.


Atmospheric Slices – Gerdec to I26

Eff

. S

ound

Spe

ed (

m/s

)

• ECMWF and G2S – almost identical up to 25km altitude

• Significant differences (~ 10 m/s) within the stratosphere

• What are the uncertainities?


Aside: Tropospheric Interaction

Large tropospheric signals for Buncefield

- due to influence of temperature inversion?


Observation Locations

Buncefield Novaky

Gerdec Chelopechene

Observed Not Observed

Wind vectorsfrom HWM

= 100m/s


Discussion / Possible Scenarios

• Can the strength of a given elevated waveguide influence the temporalevolution of the observed signals?

• If an elevated waveguide weakens – the first signals to be removed are the low angle of incidence rays – these correspond to the fast arrivals.

Documents

Analysing Multiple Explosions in Europe: